Aishan Liu

Code2Bench: Scaling Source and Rigor for Dynamic Benchmark Construction

CodeChemist: Test-Time Scaling for Low-Resource Code Generation via Functional Knowledge Transfer

RoboView-Bias: Benchmarking Visual Bias in Embodied Agents for Robotic Manipulation

Detoxifying Large Language Models via Autoregressive Reward Guided Representation Editing

Lie Detector: Unified Backdoor Detection via Cross-Examination Framework

T2V-OptJail: Discrete Prompt Optimization for Text-to-Video Jailbreak Attacks

Exploring Semantic-constrained Adversarial Example with Instruction Uncertainty Reduction

ICLShield: Exploring and Mitigating In-Context Learning Backdoor Attacks

BadRobot: Jailbreaking Embodied LLM Agents in the Physical World